HuluMed-7B: benchmark results
ZJU-AI4H's Hulu-Med, a 7B multimodal medical vision-language model covering 12 anatomical systems and 14 imaging modalities. Provider: Zhejiang University. Released 2025-10-09. Access: Open.
Aggregate Rating 1519 ± 18, rank #1977 of 10870 rated models, from 92 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MM-NeuroOnco (Closed-Ended) - Shape | 40.7 | Accuracy (%) on lesion shape regularity; five-option multipl | 100 |
| Med-R2 - Lesion Level | 58.8 | Accuracy (%) on the Med-R2 Lesion Level task (detecting and | 92.3 |
| MedQ-Deg - Artifact Degradation | 61.1 | Accuracy (%) on MedQ-Deg items whose images carry imaging-ar | 82.1 |
| MedQ-Deg - Clinical Understanding | 71 | Accuracy (%) on the Clinical Understanding questions of MedQ | 79.5 |
| MedQ-Deg - Severe Degradation (L2) | 62.2 | Accuracy (%) on the MedQ-Deg items with severe degradations | 76.9 |
| MedQ-Deg - Basic Science Reasoning | 28.6 | Accuracy (%) on the Basic Science questions of MedQ-Deg (cap | 71.8 |
| MedQ-Deg - Clean (L0) | 68.09 | Accuracy (%) on the MedQ-Deg items with the original, undegr | 71.8 |
| MedQ-Deg - Resolution and Blur Degradation | 65.8 | Accuracy (%) on MedQ-Deg items whose images carry resolution | 71.8 |
| MedQ-Deg - Mild Degradation (L1) | 67.07 | Accuracy (%) on the MedQ-Deg items with mild degradations (L | 70.5 |
| ECG-Reasoning-Benchmark (PTB-XL) - Diagnosis Given Reference Reasoning | 86.87 | Diagnosis accuracy (%) at the final decision step when the c | 69.4 |
| MedQ-Deg - Noise Degradation | 63.5 | Accuracy (%) on MedQ-Deg items whose images carry noise degr | 69.2 |
| MM-NeuroOnco (Closed-Ended) - Size | 30.3 | Accuracy (%) on lesion size; five-option multiple choice ('N | 66.7 |
Interactive version: theaggregate.ai/model?slug=hulumed-7b · How It Works · Data refreshed daily, snapshot 2026-11-12.